The is a logical extension to a technique I've been playing with whereby slit-scan panoramas are simulated using a video camera. The camera pans, then a 1 pixel wide slice is taken from each frame. These slices are appended together to make a panoramic image:
Since we only use a one pixel slice, effectively the video contains enough information for as many slit-scanned panoramas as the image has pixels in its width. By choosing a different column, you get a different result. So, I decided to make a movie by sequencing every possible panorama from one video.
This is effectively taking a 3 dimensional dataset (X,Y, time) and swapping the X and time axes.